The directory code has a dependency on the struct xfs_mount to
supply the directory block geometry. Block size, block log size,
and other parameters are pre-caclulated in the struct xfs_mount or
access directly from the superblock embedded in the struct
xfs_mount.
Extract all of this geometry information out of the struct xfs_mount
and superblock and place it into a new struct xfs_da_geometry
defined by the directory code. Allocate and initialise it at mount
time, and attach it to the struct xfs_mount so it canbe passed back
into the directory code appropriately rather than using the struct
xfs_mount.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
v5 filesystems use 512 byte inodes as a minimum, so read inodes in
clusters that are effectively half the size of a v4 filesystem with
256 byte inodes. For v5 fielsystems, scale the inode cluster size
with the size of the inode so that we keep a constant 32 inodes per
cluster ratio for all inode IO.
This only works if mkfs.xfs sets the inode alignment appropriately
for larger inode clusters, so this functionality is made conditional
on mkfs doing the right thing. xfs_repair needs to know about
the inode alignment changes, too.
Wall time:
create bulkstat find+stat ls -R unlink
v4 237s 161s 173s 201s 299s
v5 235s 163s 205s 31s 356s
patched 234s 160s 182s 29s 317s
System time:
create bulkstat find+stat ls -R unlink
v4 2601s 2490s 1653s 1656s 2960s
v5 2637s 2497s 1681s 20s 3216s
patched 2613s 2451s 1658s 20s 3007s
So, wall time same or down across the board, system time same or
down across the board, and cache hit rates all improve except for
the ls -R case which is a pure cold cache directory read workload
on v5 filesystems...
So, this patch removes most of the performance and CPU usage
differential between v4 and v5 filesystems on traversal related
workloads.
Note: while this patch is currently for v5 filesystems only, there
is no reason it can't be ported back to v4 filesystems. This hasn't
been done here because bringing the code back to v4 requires
forwards and backwards kernel compatibility testing. i.e. to
deterine if older kernels(*) do the right thing with larger inode
alignments but still only using 8k inode cluster sizes. None of this
testing and validation on v4 filesystems has been done, so for the
moment larger inode clusters is limited to v5 superblocks.
(*) a current default config v4 filesystem should mount just fine on
2.6.23 (when lazy-count support was introduced), and so if we change
the alignment emitted by mkfs without a feature bit then we have to
make sure it works properly on all kernels since 2.6.23. And if we
allow it to be changed when the lazy-count bit is not set, then it's
all kernels since v2 logs were introduced that need to be tested for
compatibility...
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
The remaining non-vectorised code for the directory structure is the
node format blocks. This is shared with the attribute tree, and so
is slightly more complex to vectorise.
Introduce a "non-directory" directory ops structure that is attached
to all non-directory inodes so that attribute operations can be
vectorised for all inodes.
Once we do this, we can vectorise all the da btree operations.
Because this patch adds more infrastructure than it removes the
binary size does not decrease:
text data bss dec hex filename
794490 96802 1096 892388 d9de4 fs/xfs/xfs.o.orig
792986 96802 1096 890884 d9804 fs/xfs/xfs.o.p1
792350 96802 1096 890248 d9588 fs/xfs/xfs.o.p2
789293 96802 1096 887191 d8997 fs/xfs/xfs.o.p3
789005 96802 1096 886903 d8997 fs/xfs/xfs.o.p4
789061 96802 1096 886959 d88af fs/xfs/xfs.o.p5
789733 96802 1096 887631 d8b4f fs/xfs/xfs.o.p6
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Lots of the dir code now goes through switches to determine what is
the correct on-disk format to parse. It generally involves a
"xfs_sbversion_hasfoo" check, deferencing the superblock version and
feature fields and hence touching several cache lines per operation
in the process. Some operations do multiple checks because they nest
conditional operations and they don't pass the information in a
direct fashion between each other.
Hence, add an ops vector to the xfs_inode structure that is
configured when the inode is initialised to point to all the correct
decode and encoding operations. This will significantly reduce the
branchiness and cacheline footprint of the directory object decoding
and encoding.
This is the first patch in a series of conversion patches. It will
introduce the ops structure, the setup of it and add the first
operation to the vector. Subsequent patches will convert directory
ops one at a time to keep the changes simple and obvious.
Just this patch shows the benefit of such an approach on code size.
Just converting the two shortform dir operations as this patch does
decreases the built binary size by ~1500 bytes:
$ size fs/xfs/xfs.o.orig fs/xfs/xfs.o.p1
text data bss dec hex filename
794490 96802 1096 892388 d9de4 fs/xfs/xfs.o.orig
792986 96802 1096 890884 d9804 fs/xfs/xfs.o.p1
$
That's a significant decrease in the instruction cache footprint of
the directory code for such a simple change, and indicates that this
approach is definitely worth pursuing further.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ben Myers <bpm@sgi.com>
Introduce a new structure xfs_trans_res to hold transaction
reservation item info per log ticket.
We also need to improve xfs_trans_resv_calc() by initializing the
log count as well as log flags for permanent log reservation.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
The struct xfs_perag has many kernel-only definitions in it,
requiring a __KERNEL__ guard so userspace can use it to. Move it to
xfs_mount.h so that it it kernel-only, and let userspace redefine
it's own version of the structure containing only what it needs.
This gets rid of another __KERNEL__ check in the XFS header files.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
xfs_mount.c is shared with userspace, but the only functions that
are shared are to do with physical superblock manipulations. This
means that less than 25% of the xfs_mount.c code is actually shared
with userspace. Move all the superblock functions to xfs_sb.c and
share that instead with libxfs.
Note that this will leave all the in-core transaction related
superblock counter modifications in xfs_mount.c as none of that is
shared with userspace. With a few more small changes, xfs_mount.h
won't need to be shared with userspace anymore, either.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
The transaction reservation size calculations is used by both kernel
and userspace, but most of the transaction code in xfs_trans.c is
kernel specific. Split all the transaction reservation code out into
it's own files to make sharing with userspace simpler. This just
leaves kernel-only definitions in xfs_trans.h, so it doesn't need to
be shared with userspace anymore, either.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
XFS_MOUNT_RETERR is going to be set at xfs_parseargs() if
mp->m_dalign is enabled, so any time we enter "if (mp->m_dalign)"
branch in xfs_update_alignment(), XFS_MOUNT_RETERR is set and so
we always be emitting a warning and returning an error.
Hence, we can remove it and get rid of a couple of redundant
check up against it at xfs_upate_alignment().
Thanks Dave Chinner for the suggestions of simplify the code
in xfs_parseargs().
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Mark Tinguely <tinguely@sgi.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Remove struct xfs_chash from struct xfs_mount as there is no user of
it nowadays.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
With the addition of CRCs, there is such a wide and varied change to
the on disk format that it makes sense to bump the superblock
version number rather than try to use feature bits for all the new
functionality.
This commit introduces all the new superblock fields needed for all
the new functionality: feature masks similar to ext4, separate
project quota inodes, a LSN field for recovery and the CRC field.
This commit does not bump the superblock version number, however.
That will be done as a separate commit at the end of the series
after all the new functionality is present so we switch it all on in
one commit. This means that we can slowly introduce the changes
without them being active and hence maintain bisectability of the
tree.
This patch is based on a patch originally written by myself back
from SGI days, which was subsequently modified by Christoph Hellwig.
There is relatively little of that patch remaining, but the history
of the patch still should be acknowledged here.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Looks the old m_inode_shrink is obsoleted as we perform inodes reclaim per AG via
m_reclaim_workqueue, this patch remove it from the xfs_mount structure if so.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Currently, we calculate the attribute set transaction
log space reservation at runtime in two parts:
1) XFS_ATTRSET_LOG_RES() which is calcuated out at mount time.
2) ((ext * (mp)->m_sb.sb_sectsize) + \
(ext * XFS_FSB_TO_B((mp), XFS_BM_MAXLEVELS(mp, XFS_ATTR_FORK))) + \
(128 * (ext + (ext * XFS_BM_MAXLEVELS(mp, XFS_ATTR_FORK))))))
which is calculated out at runtime since it depend on the given extent length in blocks.
This patch renamed XFS_ATTRSET_LOG_RES(mp) to XFS_ATTRSETM_LOG_RES(mp) to indicate
that it is figured out at mount time. Introduce XFS_ATTRSETRT_LOG_RES(mp) which would
be used to calculate out the unit of the log space reservation for one block.
In this way, the total runtime space for the given extent length can be figured out by:
XFS_ATTRSETM_LOG_RES(mp) + XFS_ATTRSETRT_LOG_RES(mp) * ext
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
CC: Dave Chinner <david@fromorbit.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Introduce a new transaction space reservation XFS_SB_LOG_RES() for
those transactions that need to modify the superblock on disk.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
CC: Dave Chinner <david@fromorbit.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Convert the calculation for end of quotaoff log space reservation
from runtime to mount time.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
CC: Dave Chinner <david@fromorbit.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Convert the calculation of quota off transaction log space reservation
from runtime to mount time.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
CC: Dave Chinner <david@fromorbit.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
The disk quota allocation log space reservation is calcuated at runtime,
this patch does it at mount time.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
CC: Dave Chinner <david@fromorbit.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
For adjusting quota limits transactions, we calculate out the log space
reservation at runtime, this patch does it at mount time.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
CC: Dave Chinner <david@fromorbit.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
The transaction log space for clearing/reseting the quota flags
is calculated out at runtime, this patch can figure it out at
mount time.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
CC: Dave Chinner <david@fromorbit.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
To separate the verifiers from iodone functions and associate read
and write verifiers at the same time, introduce a buffer verifier
operations structure to the xfs_buf.
This avoids the need for assigning the write verifier, clearing the
iodone function and re-running ioend processing in the read
verifier, and gets rid of the nasty "b_pre_io" name for the write
verifier function pointer. If we ever need to, it will also be
easier to add further content specific callbacks to a buffer with an
ops structure in place.
We also avoid needing to export verifier functions, instead we
can simply export the ops structures for those that are needed
outside the function they are defined in.
This patch also fixes a directory block readahead verifier issue
it exposed.
This patch also adds ops callbacks to the inode/alloc btree blocks
initialised by growfs. These will need more work before they will
work with CRCs.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Phil White <pwhite@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Metadata buffers that are read from disk have write verifiers
already attached to them, but newly allocated buffers do not. Add
appropriate write verifiers to all new metadata buffers.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Add a superblock verify callback function and pass it into the
buffer read functions. Remove the now redundant verification code
that is currently in use.
Adding verification shows that secondary superblocks never have
their "sb_inprogress" flag cleared by mkfs.xfs, so when validating
the secondary superblocks during a grow operation we have to avoid
checking this field. Even if we fix mkfs, we will still have to
ignore this field for verification purposes unless a version of mkfs
that does not have this bug was used.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Phil White <pwhite@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Create a new mount workqueue and delayed_work to enable background
scanning and freeing of eofblocks inodes. The scanner kicks in once
speculative preallocation occurs and stops requeueing itself when
no eofblocks inodes exist.
The scan interval is based on the new
'speculative_prealloc_lifetime' tunable (default to 5m). The
background scanner performs unfiltered, best effort scans (which
skips inodes under lock contention or with a dirty cache mapping).
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
xfs_sync.c now only contains inode reclaim functions and inode cache
iteration functions. It is not related to sync operations anymore.
Rename to xfs_icache.c to reflect it's contents and prepare for
consolidation with the other inode cache file that exists
(xfs_iget.c).
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
With the syncd functions moved to the log and/or removed, the syncd
workqueue is the only remaining bit left. It is used by the log
covering/ail pushing work, as well as by the inode reclaim work.
Given how cheap workqueues are these days, give the log and inode
reclaim work their own work queues and kill the syncd work queue.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ben Myers <bpm@sgi.com>
We don't do any data writeback from XFS any more - the VFS is
completely responsible for that, including for freeze. We can
replace the remaining caller with a VFS level function that
achieves the same thing, but without conflicting with current
writeback work.
This means we can remove the flush_work and xfs_flush_inodes() - the
VFS functionality completely replaces the internal flush queue for
doing this writeback work in a separate context to avoid stack
overruns.
This does have one complication - it cannot be called with page
locks held. Hence move the flushing of delalloc space when ENOSPC
occurs back up into xfs_file_aio_buffered_write when we don't hold
any locks that will stall writeback.
Unfortunately, writeback_inodes_sb_if_idle() is not sufficient to
trigger delalloc conversion fast enough to prevent spurious ENOSPC
whent here are hundreds of writers, thousands of small files and GBs
of free RAM. Hence we need to use sync_sb_inodes() to block callers
while we wait for writeback like the previous xfs_flush_inodes
implementation did.
That means we have to hold the s_umount lock here, but because this
call can nest inside i_mutex (the parent directory in the create
case, held by the VFS), we have to use down_read_trylock() to avoid
potential deadlocks. In practice, this trylock will succeed on
almost every attempt as unmount/remount type operations are
exceedingly rare.
Note: we always need to pass a count of zero to
generic_file_buffered_write() as the previously written byte count.
We only do this by accident before this patch by the virtue of ret
always being zero when there are no errors. Make this explicit
rather than needing to specifically zero ret in the ENOSPC retry
case.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Tested-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ben Myers <bpm@sgi.com>
The only thing the periodic sync work does now is flush the AIL and
idle the log. These are really functions of the log code, so move
the work to xfs_log.c and rename it appropriately.
The only wart that this leaves behind is the xfssyncd_centisecs
sysctl, otherwise the xfssyncd is dead. Clean up any comments that
related to xfssyncd to reflect it's passing.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ben Myers <bpm@sgi.com>
I noticed that "struct xfs_mount_args" was still declared in
"fs/xfs/xfs_mount.h". That struct doesn't even exist any more (and
is obviously not referenced elsewhere in that header file). While
in there, delete four other unneeded struct declarations in that
file.
Doing so highlights that "fs/xfs/xfs_trace.h" was relying indirectly
on "xfs_mount.h" to be #included in order to declare "struct
xfs_bmbt_irec", so add that declaration to resolve that issue.
Signed-off-by: Alex Elder <elder@inktank.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Pull second vfs pile from Al Viro:
"The stuff in there: fsfreeze deadlock fixes by Jan (essentially, the
deadlock reproduced by xfstests 068), symlink and hardlink restriction
patches, plus assorted cleanups and fixes.
Note that another fsfreeze deadlock (emergency thaw one) is *not*
dealt with - the series by Fernando conflicts a lot with Jan's, breaks
userland ABI (FIFREEZE semantics gets changed) and trades the deadlock
for massive vfsmount leak; this is going to be handled next cycle.
There probably will be another pull request, but that stuff won't be
in it."
Fix up trivial conflicts due to unrelated changes next to each other in
drivers/{staging/gdm72xx/usb_boot.c, usb/gadget/storage_common.c}
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (54 commits)
delousing target_core_file a bit
Documentation: Correct s_umount state for freeze_fs/unfreeze_fs
fs: Remove old freezing mechanism
ext2: Implement freezing
btrfs: Convert to new freezing mechanism
nilfs2: Convert to new freezing mechanism
ntfs: Convert to new freezing mechanism
fuse: Convert to new freezing mechanism
gfs2: Convert to new freezing mechanism
ocfs2: Convert to new freezing mechanism
xfs: Convert to new freezing code
ext4: Convert to new freezing mechanism
fs: Protect write paths by sb_start_write - sb_end_write
fs: Skip atime update on frozen filesystem
fs: Add freezing handling to mnt_want_write() / mnt_drop_write()
fs: Improve filesystem freezing handling
switch the protection of percpu_counter list to spinlock
nfsd: Push mnt_want_write() outside of i_mutex
btrfs: Push mnt_want_write() outside of i_mutex
fat: Push mnt_want_write() outside of i_mutex
...
Generic code now blocks all writers from standard write paths. So we add
blocking of all writers coming from ioctl (we get a protection of ioctl against
racing remount read-only as a bonus) and convert xfs_file_aio_write() to a
non-racy freeze protection. We also keep freeze protection on transaction
start to block internal filesystem writes such as removal of preallocated
blocks.
CC: Ben Myers <bpm@sgi.com>
CC: Alex Elder <elder@kernel.org>
CC: xfs@oss.sgi.com
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Rename the XFS log structure to xlog to help crash distinquish it from the
other logs in Linux.
Signed-off-by: Mark Tinguely <tinguely@sgi.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ben Myers <bpm@sgi.com>
Rename the XFS log structure to xlog to help crash distinquish it from the
other logs in Linux.
Signed-off-by: Mark Tinguely <tinguely@sgi.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ben Myers <bpm@sgi.com>
XFS_MAXIOFFSET() is just a simple macro that resolves to
mp->m_maxioffset. It doesn't need to exist, and it just makes the
code unnecessarily loud and shouty.
Make it quiet and easy to read.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
The m_maxioffset field in the struct xfs_mount contains the same
value as the superblock s_maxbytes field. There is no need to carry
two copies of this limit around, so use the VFS superblock version.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Doing background CIL flushes adds significant latency to whatever
async transaction that triggers it. To avoid blocking async
transactions on things like waiting for log buffer IO to complete,
move the CIL push off into a workqueue. By moving the push work
into a workqueue, we remove all the latency that the commit adds
from the foreground transaction commit path. This also means that
single threaded workloads won't do the CIL push procssing, leaving
them more CPU to do more async transactions.
To do this, we need to keep track of the sequence number we have
pushed work for. This avoids having many transaction commits
attempting to schedule work for the same sequence, and ensures that
we only ever have one push (background or forced) in progress at a
time. It also means that we don't need to take the CIL lock in write
mode to check for potential background push races, which reduces
lock contention.
To avoid potential issues with "smart" IO schedulers, don't use the
workqueue for log force triggered flushes. Instead, do them directly
so that the log IO is done directly by the process issuing the log
force and so doesn't get stuck on IO elevator queue idling
incorrectly delaying the log IO from the workqueue.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Now that we write back all metadata either synchronously or through
the AIL we can simply implement metadata freezing in terms of
emptying the AIL.
The implementation for this is fairly simply and straight-forward:
A new routine is added that asks the xfsaild to push the AIL to the
end and waits for it to complete and send a wakeup. The routine will
then loop if the AIL is not actually empty, and continue to do so
until the AIL is compeltely empty.
We keep an inode reclaim pass in the freeze process to avoid having
memory pressure have to reclaim inodes that require dirtying the
filesystem to be reclaimed after the freeze has completed. This
means we can also treat unmount in the exact same way as freeze.
As an upside we can now remove the radix tree based inode writeback
and xfs_unmountfs_writesb.
[ Dave Chinner:
- Cleaned up commit message.
- Added inode reclaim passes back into freeze.
- Cleaned up wakeup mechanism to avoid the use of a new
sleep counter variable. ]
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
The new concurrency managed workqueues are cheap enough that we can create
per-filesystem instead of global workqueues. This allows us to remove the
trylock or defer scheme on the ilock, which is not helpful once we have
outstanding log reservations until finishing a size update.
Also allow the default concurrency on this workqueues so that I/O completions
blocking on the ilock for one inode do not block process for another inode.
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ben Myers <bpm@sgi.com>
Change xfs_sb_from_disk() interface to take a mount pointer
instead of a superblock pointer.
This is to print mount point specific error messages in future
fixes.
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Ben Myers <bpm@sgi.com>
The delaylog mode has been the default for a long time, and the nodelaylog
option has been scheduled for removal in Linux 3.3. Remove it and code
only used by it now that we have opened the 3.3 window.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
Remove the second parameter to xfs_sb_count() since all callers of
the function set them.
Also, fix the header comment regarding it being called periodically.
Signed-off-by: Chandra Seetharaman <sekharan@us.ibm.com>
Signed-off-by: Alex Elder <aelder@sgi.com>
Now that we have reliably tracking of deleted extents in a
transaction we can easily implement "online" discard support
which calls blkdev_issue_discard once a transaction commits.
The actual discard is a two stage operation as we first have
to mark the busy extent as not available for reuse before we
can start the actual discard. Note that we don't bother
supporting discard for the non-delaylog mode.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Alex Elder <aelder@sgi.com>
Background inode reclaim needs to run more frequently that the XFS
syncd work is run as 30s is too long between optimal reclaim runs.
Add a new periodic work item to the xfs syncd workqueue to run a
fast, non-blocking inode reclaim scan.
Background inode reclaim is kicked by the act of marking inodes for
reclaim. When an AG is first marked as having reclaimable inodes,
the background reclaim work is kicked. It will continue to run
periodically untill it detects that there are no more reclaimable
inodes. It will be kicked again when the first inode is queued for
reclaim.
To ensure shrinker based inode reclaim throttles to the inode
cleaning and reclaim rate but still reclaim inodes efficiently, make it kick the
background inode reclaim so that when we are low on memory we are
trying to reclaim inodes as efficiently as possible. This kick shoul
d not be necessary, but it will protect against failures to kick the
background reclaim when inodes are first dirtied.
To provide the rate throttling, make the shrinker pass do
synchronous inode reclaim so that it blocks on inodes under IO. This
means that the shrinker will reclaim inodes rather than just
skipping over them, but it does not adversely affect the rate of
reclaim because most dirty inodes are already under IO due to the
background reclaim work the shrinker kicked.
These two modifications solve one of the two OOM killer invocations
Chris Mason reported recently when running a stress testing script.
The particular workload trigger for the OOM killer invocation is
where there are more threads than CPUs all unlinking files in an
extremely memory constrained environment. Unlike other solutions,
this one does not have a performance impact on performance when
memory is not constrained or the number of concurrent threads
operating is <= to the number of CPUs.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Alex Elder <aelder@sgi.com>
On of the problems with the current inode flush at ENOSPC is that we
queue a flush per ENOSPC event, regardless of how many are already
queued. Thi can result in hundreds of queued flushes, most of
which simply burn CPU scanned and do no real work. This simply slows
down allocation at ENOSPC.
We really only need one active flush at a time, and we can easily
implement that via the new xfs_syncd_wq. All we need to do is queue
a flush if one is not already active, then block waiting for the
currently active flush to complete. The result is that we only ever
have a single ENOSPC inode flush active at a time and this greatly
reduces the overhead of ENOSPC processing.
On my 2p test machine, this results in tests exercising ENOSPC
conditions running significantly faster - 042 halves execution time,
083 drops from 60s to 5s, etc - while not introducing test
regressions.
This allows us to remove the old xfssyncd threads and infrastructure
as they are no longer used.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Alex Elder <aelder@sgi.com>
All of the work xfssyncd does is background functionality. There is
no need for a thread per filesystem to do this work - it can al be
managed by a global workqueue now they manage concurrency
effectively.
Introduce a new gglobal xfssyncd workqueue, and convert the periodic
work to use this new functionality. To do this, use a delayed work
construct to schedule the next running of the periodic sync work
for the filesystem. When the sync work is complete, queue a new
delayed work for the next running of the sync work.
For laptop mode, we wait on completion for the sync works, so ensure
that the sync work queuing interface can flush and wait for work to
complete to enable the work queue infrastructure to replace the
current sequence number and wakeup that is used.
Because the sync work does non-trivial amounts of work, mark the
new work queue as CPU intensive.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Alex Elder <aelder@sgi.com>
Currently the size of the speculative preallocation during delayed
allocation is fixed by either the allocsize mount option of a
default size. We are seeing a lot of cases where we need to
recommend using the allocsize mount option to prevent fragmentation
when buffered writes land in the same AG.
Rather than using a fixed preallocation size by default (up to 64k),
make it dynamic by basing it on the current inode size. That way the
EOF preallocation will increase as the file size increases. Hence
for streaming writes we are much more likely to get large
preallocations exactly when we need it to reduce fragementation.
For default settings, the size of the initial extents is determined
by the number of parallel writers and the amount of memory in the
machine. For 4GB RAM and 4 concurrent 32GB file writes:
EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET TOTAL
0: [0..1048575]: 1048672..2097247 0 (1048672..2097247) 1048576
1: [1048576..2097151]: 5242976..6291551 0 (5242976..6291551) 1048576
2: [2097152..4194303]: 12583008..14680159 0 (12583008..14680159) 2097152
3: [4194304..8388607]: 25165920..29360223 0 (25165920..29360223) 4194304
4: [8388608..16777215]: 58720352..67108959 0 (58720352..67108959) 8388608
5: [16777216..33554423]: 117440584..134217791 0 (117440584..134217791) 16777208
6: [33554424..50331511]: 184549056..201326143 0 (184549056..201326143) 16777088
7: [50331512..67108599]: 251657408..268434495 0 (251657408..268434495) 16777088
and for 16 concurrent 16GB file writes:
EXT: FILE-OFFSET BLOCK-RANGE AG AG-OFFSET TOTAL
0: [0..262143]: 2490472..2752615 0 (2490472..2752615) 262144
1: [262144..524287]: 6291560..6553703 0 (6291560..6553703) 262144
2: [524288..1048575]: 13631592..14155879 0 (13631592..14155879) 524288
3: [1048576..2097151]: 30408808..31457383 0 (30408808..31457383) 1048576
4: [2097152..4194303]: 52428904..54526055 0 (52428904..54526055) 2097152
5: [4194304..8388607]: 104857704..109052007 0 (104857704..109052007) 4194304
6: [8388608..16777215]: 209715304..218103911 0 (209715304..218103911) 8388608
7: [16777216..33554423]: 452984848..469762055 0 (452984848..469762055) 16777208
Because it is hard to take back specualtive preallocation, cases
where there are large slow growing log files on a nearly full
filesystem may cause premature ENOSPC. Hence as the filesystem nears
full, the maximum dynamic prealloc size іs reduced according to this
table (based on 4k block size):
freespace max prealloc size
>5% full extent (8GB)
4-5% 2GB (8GB >> 2)
3-4% 1GB (8GB >> 3)
2-3% 512MB (8GB >> 4)
1-2% 256MB (8GB >> 5)
<1% 128MB (8GB >> 6)
This should reduce the amount of space held in speculative
preallocation for such cases.
The allocsize mount option turns off the dynamic behaviour and fixes
the prealloc size to whatever the mount option specifies. i.e. the
behaviour is unchanged.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Dave Chinner